Bogdan Suvernev

My research focuses on large language models, reinforcement learning, and AI for education. I am particularly interested in post-training methods for LLMs, including PPO, DPO, and GRPO, as well as retrieval-augmented generation (RAG), efficient model training, and reasoning capabilities. My recent work explores Socratic tutoring with LLMs, reinforcement learning for improving educational dialogue systems, and scalable distributed training of foundation models. I also have experience building diffusion models, reinforcement learning agents, and high-performance AI systems from scratch in PyTorch. I am interested in collaborating on research related to LLM alignment, reasoning, AI agents, and open-source foundation models.